CN100479527C

CN100479527C - Method for selecting predicting mode within frame

Info

Publication number: CN100479527C
Application number: CN 200410006340
Authority: CN
Inventors: 邹立建; 黄庆明; 高文
Original assignee: National Source Coding Center Digital Audio And Video Frequency Technology (beijing) Co Ltd
Current assignee: UNITED XINYUAN DIGITAL AUDIO V
Priority date: 2004-02-26
Filing date: 2004-02-26
Publication date: 2009-04-15
Anticipated expiration: 2024-02-26
Also published as: CN1662066A

Abstract

The method includes following steps: partitioning image to be decoded; analyzing texture feature of blocks of image to be decoded, especially analyzing regional directional feature of texture in image; selecting prediction mode in frame, where the predicted direction is nearest to ideal prediction direction, is as optimal prediction mode in the block. Advantages are: without need of carrying out inverse calculation for each probability, selecting optimal prediction mode quickly, reducing amount of calculation and in favor of realizing real time video compress system.

Description

A kind of intra prediction mode choosing method

Technical field

The present invention relates to a kind of video-frequency compression method, relate in particular to the method for using method of intra-prediction that single-frame images is compressed.

Background technology

At present, some video compression standard is introduced intra-frame prediction method to eliminate the spatial redundancy information in the single-frame images, as video compression standard H.264, this method is a frame sign that the image division of M * N is piece (P, Q=4,8 of a plurality of P * Q earlier, 16 ...), as shown in Figure 1, carry out conversion to first earlier in order then, quantize, inverse quantization, inverse transformation obtains first reconstructed block, when second piece encoded, predict according to first reconstructed block earlier, as shown in Figure 2, carry out the horizontal direction prediction, and then the residual block of original picture block and predicted picture piece is encoded; For the most several piece in the two field picture, doping this piece image according to neighbor has a variety of predictive modes, and its difference is the prediction direction difference, as shown in Figure 3, the value of x11-x44 pixel can be according to this image block neighbor A-I in the image block, and O-V predicts out.Best predictive mode can make prediction piece and original block similarity degree maximum, thereby makes residual block coding back data volume minimum.

In the prior art, the general method that adopts of infra-frame prediction may further comprise the steps:

A) for piece to be encoded, according to piece encoded and that rebuild, which pixel is available around judging, thereby which predictive mode decision has available;

B) to a kind of predictive mode, predict, obtain predicting piece;

C) the residual image piece of calculating original picture block and predicted picture piece is preserved;

E) according to all pixel absolute values in the residual image piece and (SAD), from all predictive modes, the predictive mode of choosing its SAD minimum is an optimum prediction mode;

This method need be calculated each predictive mode, and required amount of calculation is big, is unsuitable for the real-time coding system.

Summary of the invention

The technical problem to be solved in the present invention is to overcome existent method need calculate each possible predictive mode, and the defective that the amount of calculation of bringing is big and calculating is loaded down with trivial details, in order to help the realization of real-time video compressibility, thereby provide a kind of the need each possible predictive mode to be calculated the intra prediction mode choosing method.

In order to solve the problems of the technologies described above, the invention provides a kind of intra prediction mode choosing method, may further comprise the steps:

A) treat coded image and carry out piecemeal, by the coded picture block compositing area of cutting apart;

B) each regional grain direction feature in textural characteristics, the especially image of analysis image block to be encoded is come out these direction characters mark in image; Wherein, ganmma controller gradient analysis method and Fourier Time-frequency Spectrum Analysis method are adopted in the analysis of described image texture characteristic;

C) for certain image block to be encoded, according to the grain direction of region, determine the prediction direction that it is desirable, this desirable prediction direction is exactly the grain direction of image block to be encoded region;

D) according to compression algorithm, calculate the direction value direc[d of all predictive mode correspondences]; From the direction value of all predictive modes, choose value bestdirec near this image texture direction; The pairing predictive mode of the value bestdirec of this image texture direction is the optimum prediction mode of this image block.

In such scheme, according to video-frequency compression method image is carried out piecemeal in the step a).

In such scheme, in the described step b), when the data of a data block are carried out texture analysis, obtain the grain direction value of this data block.

In such scheme, step c) may further comprise the steps:

C1) data of a data block are carried out texture analysis, obtain the grain direction value of this data block;

C2) according to compression algorithm, calculate the direction value direc[d of all predictive mode correspondences].

In such scheme, step d) may further comprise the steps:

D1) from the direction value of all predictive modes, choose value bestdirec near this image texture direction;

D2) the pairing predictive mode of bestdirec is the optimum prediction mode of this image block.

The invention has the advantages that: intra prediction mode choosing method provided by the invention, do not need each possible predictive mode is calculated, optimum prediction mode that can fast selecting I picture piece has reduced amount of calculation widely, helps the realization of real-time video compressibility.

Description of drawings

Fig. 1 is a prior art image division code pattern of the present invention;

Fig. 2 is horizontal direction prognostic chart in the prior art of the present invention;

Fig. 3 selects figure for predictive mode in the prior art of the present invention;

Fig. 4 is the inventive method flow chart;

Fig. 5 is an embodiment of the invention image texture characteristic directional diagram;

Fig. 6 is the method flow diagram of the embodiment of the invention.

Embodiment

Describe technical scheme of the present invention in detail below in conjunction with attached Figure 4 and 5.

Fig. 4 is a method flow diagram of the present invention, shows a kind of method flow of directly choosing optimum prediction mode in certain image block frame according to image texture features.

As shown in Figure 5, image texture features, particularly its direction character have reflected the similitude of image on this grain direction.

As shown in Figure 4, a kind of fast frame inner estimation mode choosing method may further comprise the steps:

Step 100 according to the requirement of video compression algorithm, is treated coded image and is carried out piecemeal, and for example the MPEG-4 algorithm requires to be divided into 8 * 8 block of pixels;

Step 110 is analyzed each regional grain direction feature in textural characteristics, the especially image of image block to be encoded, these direction characters mark in image is come out, as shown in Figure 5; The analysis of image texture characteristic can be according to the ripe algorithm of present image processing, as gradient of image and gray scale analytic approach, Fourier spectrum analytic approach etc.;

Step 120 for certain image block to be encoded, according to the grain direction of region, is determined desirable prediction direction; As 6 image blocks in the upper left corner among Fig. 5, the grain direction of region is horizontal direction, and its perfect forecast direction is exactly a horizontal direction so;

Step 130 is chosen its prediction direction near the intra prediction mode of the perfect forecast direction optimal prediction modes as this piece, is horizontal forecast pattern 1 as 6 image block optimum prediction modes in the upper left corner among Fig. 5.

As shown in Figure 6, implementing a kind of fast frame inner estimation mode choosing method may further comprise the steps:

Step 200 is handled beginning;

Step 210 is read in data from a pending two field picture, and it is required to be divided into little image block according to algorithm, and for example 8*8 is kept at data set block[B] [8] [8], in this step, algorithm can be the MPEG-4 algorithm;

Step 220 reads the data ONE BLOCK[8 of a data block in order] [8];

Step 230 is carried out texture analysis to the data of a data block, obtains the grain direction value of this data block;

Step 240 according to compression algorithm, is calculated the direction value direc[d of all predictive mode correspondences];

Step 250 is chosen the value bestdirec near this image texture direction from the direction value of all predictive modes;

Step 260, the pairing predictive mode of bestdirec is the optimum prediction mode of this image block;

Is step 270 handled all image blocks? if carry out next step; Be not worth execution in step 220;

Step 280, processing finishes.

In sum, intra prediction mode choosing method provided by the invention does not need each possible predictive mode is calculated, and optimum prediction mode that can fast selecting I picture piece reduces amount of calculation, helps the realization of real-time video compressibility.

Claims

1, a kind of intra prediction mode choosing method is characterized in that may further comprise the steps:

A). treat coded image and carry out piecemeal, by the coded picture block compositing area of cutting apart;

B). analyze the textural characteristics of image block to be encoded, comprise each regional grain direction feature in the image, these direction characters mark in image is come out; Wherein, ganmma controller gradient analysis method or fourier spectrum analytic approach are adopted in the analysis of described image texture characteristic;

C). for certain image block to be encoded, according to the grain direction of region, determine the prediction direction that it is desirable, this desirable prediction direction is exactly the grain direction of image block to be encoded region;

D). according to compression algorithm, calculate the direction value direc[d of all predictive mode correspondences]; From the direction value of all predictive modes, choose value bestdirec near this image texture direction; The pairing predictive mode of the value bestdirec of this image texture direction is the optimum prediction mode of this image block.

2, a kind of intra prediction mode choosing method as claimed in claim 1 is characterized in that, according to video-frequency compression method image is carried out piecemeal in the described step a).

3, a kind of intra prediction mode choosing method as claimed in claim 1 is characterized in that, in the described step b), when the data of a data block are carried out texture analysis, obtains the grain direction value of this data block.